Overview of MELCOR Code Activities in IBRAE RAN

Size: px

Start display at page:

Download "Overview of MELCOR Code Activities in IBRAE RAN"

Kelly Montgomery
5 years ago
Views:

Institute (IBRAE) Overview of MELCOR Code Activities in IBRAE RAN I.

1 РОССИЙСКАЯ АКАДЕМИЯ НАУК Институт проблем безопасного развития атомной энергетики RUSSIAN ACADEMY OF SCIENCES Nuclear Safety Institute (IBRAE) Overview of MELCOR Code Activities in IBRAE RAN I. Drobyshevskaya N. Mosunova A. Gorobets VUJE a.s. Bratislava, Slovak Republic

2 Outline MELCOR 2.1 code performance improvement strategies Optimization and refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization COR and RN1 packages overlap mode MELOR 2.1 code verification and validation on the Lower Head Failure (LHF) experiments 2

3 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization COR and RN1 packages overlap mode 3

4 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization COR and RN1 packages overlap mode Global data have been replaced with local copies. Testing of these changes has shown noticeable speedup on the tests. Test case Rev. 5250, sec Rev. 5441, sec Speedup, % BWR/Mark I SBO PWR LBLOCA BWR/Mark III SBO

5 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization Realized for all MELCOR packages Decrease of total CPU time by 1 2 % COR and RN1 packages overlap mode Original version Copying Old = New before each time step. Copying New = Old if fallback Version with swapping Two pointers pold and pnew Two objects Old and New pold => Old pnew => New 5

6 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization CVH and RN1 packages have been parallelized Decrease of total CPU time by 5 10 % COR and RN1 packages overlap mode Temperature of the cladding for BWR/Mark I SBO test case Temperature of the cladding for PWR LBLOCA test case 6

7 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization COR and RN1 packages overlap mode Base linear solver BiCG replaced by its modification BiCGSTAB It has faster and smoother convergence than the original BiCG Solver speedup about 3 4 times Test case BiCG, sec BiCGSTAB, sec Speedup BWR/Mark I SBO 1191,16 348,92 3,41 BWR/Mark III LBLOCA 3776,63 857,19 4,4 7

8 MELCOR 2.1 Performance Improvement Refactoring Swapping algorithms Loop-based parallelization Numerical solvers modernization COR and RN1 packages overlap mode The distribution of COR package is not appropriate for efficient loop-based parallelization COR and RN1 packages both took comparable CPU time (generally at a ratio of 3 to 1) CVH CVH CVH CVH COR COR COR RN1 COR RN1 RN1 RN1 8

9 COR and RN1 Overlap The distribution of COR package is not appropriate for efficient loop-based parallelization COR and RN1 packages both took comparable CPU time (generally at a ratio of 3 to 1) CVH CVH CVH CVH COR COR COR RN1 COR RN1 RN1 RN1 9

10 Changes in the Algorithm t n layer COR Data XMRCOR_Tot XMRCOR_Rad Other fields RN1 Data XMRCOR_Tot XMRCOR_Rad t n-1 layer RN1 Data Other fields 10

11 Changes in the Algorithm t n layer COR Data XMRCOR_Tot XMRCOR_Rad Other fields t n-1 layer RN1 Data XMRCOR_Tot XMRCOR_Rad Other fields 11

12 Sensitivity Coefficient SC Criteria for activation and deactivation of overlap mode of RN1 and COR packages (1) - The flag for used mode Value Comment 0 The original mode is used COR and RN1 take values from the same temporal layer 1 2 The new COR values used on the next step by RN1 package, but COR and RN1 packages are calculated subsequently The new COR values used on the next step by RN1 package, COR and RN1 packages are calculated simultaneously 12

13 Performance Testing Release configuration COR and RN1 overlap mode Test case BWR/Mark I SBO BWR/Mark III SBO BWR/Mark III LBLOCA CPU time, sec/ hours SC = 0 SC = 1 SC =

14 Calculation Results Physical results for base version and version with modified algorithm in RN1 package are of the same quality The modified version run in sequential and parallel mode gives identical results for most of the tests Cladding temperature for the BWR/Mark III LBLOCA test case Total radioactive mass release for the BWR/Mark III SBO test case 14

15 Lower Head Failure Experiments Purpose Experimental study of the vessel creep and deformation process Experimental results utilization to develop and validate analytical models Conducted in 1998 in Sandia National Laboratory Series of 8 experiments distinguished by Spatial temperature/heat flux distribution Pressure Reactor vessel structure elements and construction features on RPV deformation and failure 15

16 Test Matrix of LHF Experiments Tests Heat Flux Distribution Structure Elements Pressure LHF-1 Uniform 10 MPa LHF-2 Center-peaked 10 MPa LHF-3 Edged-peaked 10 MPa LHF-4 Uniform Penetrations 10 MPa LHF-5 Edge-peaked Penetrations 10 MPa with transient LHF-6 Uniform Weldment 10 MPa LHF-7 Uniform 5 MPa LHF-8 Edge-peaked 10 MPa 16

$material weakening at elevated temperatures Creep is calculated based on a Larson- Miller parameter and a life-fraction rule ε pl ( t + t ) = ε$

17 MELCOR 2.1 Model for Lower Head Failure Creep-rupture failure of a lower head segment occurs, in response to mechanical loading under conditions of material weakening at elevated temperatures Creep is calculated based on a Larson- Miller parameter and a life-fraction rule ε pl ( t + t ) = ε pl ( t ) Zero-dimensional option: effective stress ( P + ρ g z e = d σ 2 Ro - R d 2 i ) R One-dimensional option: predicts the stressstrain distribution through the lower head, and treats stress redistribution from both thermal strain and material property degradation 2 i t t R 17

18 Mechanical Properties of the Vessel Steel 18

19 LHF Test Nodalization Scheme CV600 Environment 19

20 LHF-4 Experiment (Boundary Conditions) 20

21 LHF-3 Experiment (Boundary Conditions) 21

22 Calculation Results for LHF-4 Experiment Displacement, 90 degrees Displacement, 45 degrees 22

23 Calculation Results for LHF-3 Experiment Displacement, 90 degrees Displacement, 30 degrees Displacement, cm 1,8 1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 Calculation 1D Calculation 0D Exp. data, 30 deg Time, s 23

24 Sensitivity Study for LHF-5 Experiment Displacement, 90 degrees Displacement, 90 degrees 24

25 Conclusions MELCOR 2.1 code activities in IBRAE RAN are presented Code performance improvement Code validation and verification Code performance improvement New approach of code performance improvement: COR and RN1 packages overlap - decrease of CPU time by 10-30% for NPP calculations Code refactoring - decrease of CPU time by 10-15% for NPP calculations Code verification on LHF experiments Good agreement of MELCOR code calculation results with experimental data for the failure time and the LH pole displacement Using of 1D option to calculate stress-strain distribution through the lower head is recommended Special attention should be paid to set the correct values for material properties MELCOR model underestimates the pole displacement for the experiments with the LH uniform heating

Expansion of the model basis in MELCOR

Expansion of the model basis in MELCOR Philipp Dietrich KIT University of the State of Baden-Württemberg and National Research Center of the Helmholtz Association www.kit.edu Outline Motivation Adding